TrueVoice API

Know when a voice is AI.

TrueVoice listens to a recording and tells you whether the speaker is a real person or a synthetic voice.

Supports

English
Arabic
Spanish
French
Hindi
Portuguese

For developers

How the API works

You post an audio or video file of up to 60 seconds. The response labels the whole clip as human or AI, plus a separate label for every four seconds, which shows where a synthetic voice starts and stops in a mixed recording.

curl https://api.advancedinitiatives.com/v1/score \
  -H "Authorization: Bearer $TRUEVOICE_KEY" \
  -F file=@call.wav

import requests

r = requests.post("https://api.advancedinitiatives.com/v1/score",
    headers={"Authorization": f"Bearer {key}"},
    files={"file": open("call.wav", "rb")})
print(r.json()["label"])   # "human" or "ai"

const body = new FormData();
body.append("file", file);
const r = await fetch("https://api.advancedinitiatives.com/v1/score", {
  method: "POST", headers: { Authorization: `Bearer ${key}` }, body,
});
const { label, segments } = await r.json();

Response

{
  "label": "ai",
  "band": "clear",
  "seconds": 12.4,
  "segments": [
    { "start": 0, "end": 4,  "label": "human" },
    { "start": 4, "end": 8,  "label": "ai" },
    { "start": 8, "end": 12.4, "label": "ai" }
  ]
}

Clip length

Anything from 1 to 60 seconds. Split longer recordings and send the pieces.

Formats

WAV, MP3, M4A, FLAC, OGG, WebM, and the audio track of an MP4 video.

Close calls

Sections near the decision threshold come back marked as uncertain.

Privacy

We delete the audio as soon as it has been checked.

Request API access

Fill in the form and we will email you an API key and setup instructions.

Questions first? Write to info@advancedinitiatives.com.